Why is my WD1500ADFD Raptor performing so slowly??

[lonkny]

Hi all,

I have a frustrating issue that I'm trying to resolve. I have 2 10K WD Raptor drives - a WD1500ADFD-00NLR5 (FW 21.07) and a WD1500HLFS-01G7U0 (FW 04.04) in a Dell Dimension 9200 (HTML specs here) running Windows XP Pro SP3 x86. I would like to use these Raptors in a RAID0 array to use as a data drive for pro audio sample libraries, audio/video editing, etc. I realize this provides no redundancy so will be backing up this data regularly. I need the performance of RAID0 for video & audio editing functions/apps.

The Intel P965 motherboard uses Intel ICH8R and BIOS (latest version) is in RAID mode. I currently have each drive in the first 3 SATA slots (including System drive) and no RAID array is currently configured. Here is where the problems start - I have benchmarked the 2 150GB Raptors and consistently receive similar results for each... I will post a typical HD Tach result for each drive below:

WD1500HLFS 150GB 10K VelociRaptor:

WD1500ADFD 150GB 10K Raptor:

Why is the WD1500ADFD so much slower than the first drive?? I realize it is an older model, but has the performance increased that dramatically for Raptors? I would like to combine these 2 drives in a RAID0 config, but am very reluctant at this point because of the huge differences in performance. Also, why such large dips at the beginning of the test (down to 10MB/s at one point!)? These huge dips in performance were occurring in every test on this drive. Hoping these dips in performance may provide a clue as to what the issue is...

Just to be sure, here are some of the other things I have tried to ensure it wasn't an issue with my setup. I have a VIA VT6421 addon PCI RAID card (Rosewill RC-212) and connected both drives to this addon card instead - same test results. I then tried connecting the drives to different SATA ports on the motherboard - same results. I then tried setting the jumper on the WD1500ADFD to turn on 1.5GB/sec - see Jumper Setting #4 on this page - same results. Finally, I reformatted the drive as NTFS (default allocation unit size) and did not use the "Quick Format" option to ensure a clean, completely formatted disk. After all these steps, the drive still produces the same results in both HD Tach and HDTune.

My questions are:

* Why does the HLFS have 7.1ms Random Access time while ADFD has 8.1ms?
* Why does the HLFS have 251.4MB/s burst speed while ADFD has 132.1MB/s (slower than ATA UDMA6!!)
* Why does the HLFS have relatively consistent read speed while the ADFD experiences several severe drops in speed, especially towards the beginning of each test. This was the same for HD Tach and HDTune.
* As mentioned, the drive has FW v21.07 - is there an updated version available from WD tech phone support? There is none available on the support page for the drive.
* Should I leave the jumper for 1.5Gb/s enabled? Should I also enable the jumper for Spread Spectrum Clocking (see #2 on Jumper page)? Currently this jumper is not enabled.

Is there anything else I can try to improve the performance of my WD1500ADFD drive?

Hoping someone out there can give me a hand with this, as I am at my wit's end! Thanks for any help or info you can provide!!

All the best,
LonkNY

 

[cjl]

1)The Velociraptor has a faster random access time primarily because it is a 2.5" drive, while the Raptor is 3.5". This means the servo mechanism is smaller and can move faster.

2) The burst speed on the Raptor is slow because it is limited to 1.5Gbit SATA. This gives a maximum transfer rate of 150MB/s. Velociraptors use 3Gbit SATA, giving a maximum rate of 300MB/s.

3) Is the velociraptor freshly formatted? Is the Raptor your current OS drive? That would explain the issue. Usually, large dips means that something else is trying to access it, and this is common with an OS drive.

4) Very few hard drives have user-upgradeable firmware.

5) Honestly, the 1.5Gbit jumper shouldn't matter at all. As for spread spectrum, I wouldn't worry about it.

 

[lonkny]

Thanks for the reply... to answer #3, both drives are freshly formatted and neither are the current OS drive. They are both recently added to the box and are currently setup as standalone drives. Nothing should be accessing either drive. I also disabled realtime Antivirus engine while running these tests just in case.

In your opinion, for a 150Gb/s WD1500ADFD that's a couple years old, are the numbers above considered slow for this drive? They seem to be, but may they aren't... I guess if these are the numbers, then it isn't worthwhile combining these 2 drives in a RAID0 array since the older Raptor would slow the new one down so much...

If this is the case, I guess it's back to the drawing board for me! Plz let me know, thanks!

 

[Pointertovoid]

Answer 1) is just plain false. False diameters.

Answer 2) as well. Throughput is related to platter performance, not interface.

In answer 3), formatting is irrelevant. Hd Tach accesses the drives independently of any format or data on it. It doesn't rely on the file system and works even without formatting nor partitioning the disk. It requests I/O operations on sector numbers, that's why only administrators can run this software.

Performance differs: yes. Research labs get paid for that.

Dips in the curve: they're due to other tasks AND to the horrible Hd Tach that's so sensitive to them. Winbench, h2benchw are better to measure contiguous read. Atto gives reasonable values as well.

Raid-0: for big files, still sensible. But then, Raptors aren't a cheap choice! For 45 euro each, you get 7200rpm disks with one 500GB platter (Seagate 7200.12, Samsung F3, Hitachi 7k1000c (not b), and more at WD and Maxtor) which deliver 134MB/s contiguous read. You should rather keep the Raptors for the OS and applications. And different performance aren't an issue; you'll get twice the slowest contiguous read. Take a huge stripe size with recent disks: about 256kB for Raptors, 512kB for 7200rpm.

 

[notty22]

Answer 1) is just plain false. False diameters.

Answer 2) as well. Throughput is related to platter performance, not interface.

In answer 3), formatting is irrelevant. Hd Tach accesses the drives independently of any format or data on it. It doesn't rely on the file system and works even without formatting nor partitioning the disk. It requests I/O operations on sector numbers, that's why only administrators can run this software.

Performance differs: yes. Research labs get paid for that.

Dips in the curve: they're due to other tasks AND to the horrible Hd Tach that's so sensitive to them. Winbench, h2benchw are better to measure contiguous read. Atto gives reasonable values as well.

Raid-0: for big files, still sensible. But then, Raptors aren't a cheap choice! For 45 euro each, you get 7200rpm disks with one 500GB platter (Seagate 7200.12, Samsung F3, Hitachi 7k1000c (not b), and more at WD and Maxtor) which deliver 134MB/s contiguous read. You should rather keep the Raptors for the OS and applications. And different performance aren't an issue; you'll get twice the slowest contiguous read. Take a huge stripe size with recent disks: about 256kB for Raptors, 512kB for 7200rpm.

THIS IS THE WRONG ADVICE: ^^^^^^^^

Cj was correct with his answers.
Go over to wikipedia, they have the history of the raptor and its newer cousin the Velociraptor. I'm a original owner of the raptor. Its a 3.5 drive in a 3.5 enclosure. The velocoraptor is a 2.5 inch drive in a 3.5 housing where the extra housing is a giant heat sink
http://www.newegg.com/Product/ImageGallery.aspx?CurImage=22-136-296-S01&SCList=22-136-296-S01%2c22-136-296-S02%2c22-136-296-S03%2c22-136-296-S04%2c22-136-296-S05%2c22-136-296-S06&S7ImageFlag=2&Item=N82E16822136296&Depa=0&WaterMark=1&Description=Western%20Digital%20VelociRaptor%20WD1500HLFS%20150GB%2010000%20RPM%20SATA%203.0Gb%2fs%203.5%22%20Internal%20Hard%20Drive%20-Bare%20Drive
He is also correct with answers 2 + 3

edit : #2 sequential read speed is partly limited to the interface efficiency and the platter size diff
between the two drives. They both spin at 10,000 rpm,same capacity but at the outer ring of a 2.5 platter vs the outer ring of a 3.5 inch platter helps account for the higher throughput. My 10,000 rpm 37,g drive only transfers 56mb a sec, its the original raptor.
HD tune pro 4.1
37 g 10,000 rpm raptor
HD Tune Pro: WDC WD360GD-00FNA0 Extra Tests
Random seek 111 IOPS 9.0 ms 0.054 MB/s
Butterfly seek 103 IOPS 9.7 ms 0.050 MB/s
Random seek / size 64 KB 99 IOPS 10.1 ms 1.529 MB/s
Random seek / size 8 MB 9 IOPS 112.6 ms 36.021 MB/s
Sequential read outer 880 IOPS 1.1 ms 54.982 MB/s
Sequential read middle 783 IOPS 1.3 ms 48.944 MB/s
Sequential read inner 542 IOPS 1.8 ms 33.886 MB/s
Burst rate 1464 IOPS 0.7 ms 91.480 MB/s

1 TB Caviar Black 7200 rpm drive
HD Tune Pro: WDC WD1001FALS-00J7B1 Extra Tests
Random seek 82 IOPS 12.2 ms 0.040 MB/s
Butterfly seek 69 IOPS 14.6 ms 0.034 MB/s
Random seek / size 64 KB 81 IOPS 12.4 ms 1.239 MB/s
Random seek / size 8 MB 16 IOPS 61.5 ms 65.915 MB/s
Sequential read outer 1741 IOPS 0.6 ms 108.791 MB/s
Sequential read middle 1492 IOPS 0.7 ms 93.244 MB/s
Sequential read inner 870 IOPS 1.2 ms 54.346 MB/s
Burst rate 2651 IOPS 0.4 ms 165.684 MB/s

 

[lonkny]

Thanks guys for all of the different opinions! You are correct in that the Velociraptor is a 2.5" in 3.5" casing/heatsink, while the Raptor is a 3.5" drive...

So it sounds like my numbers may not be far off from what they should be then... and that the Velociraptor is indeed a much faster drive than the Raptor...

I do actually have another 160GB 10K drive that came with the Dell that I could pair with the Raptor in a 300GB RAID0 for large pro audio/video files... the possible issue with this is I believe it only has 8MB cache whereas the WD1500ADFD Raptor has 16MB cache - would this be OK to pair these 2 drives in a RAID0? I could then image the OS to the Velociraptor and have the best of both worlds (fast Velociraptor handling OS/Apps and not being slowed down by pairing it with the older Raptor, and instead pairing the old Raptor w/the other 10K 8MB cache 160GB drive in RAID0 - I would lose 10GB on one of the drives but WTF)... what does everyone think of this?! I think this would probably be the best plan... thoughts?

 

[cjl]

Cache size doesn't matter very much. The RAID with an old Raptor and the other old 160GB would work though - are you sure the 160 is SATA? I would have expected it to be SCSI or SAS, but as long as it is SATA, that plan sounds like a good way to go.

 

[lonkny]

OK great, thx... thanks all - this is a great forum/site.

 

[lonkny]

Best answer selected by lonkny.